Make changes to pre-existing AI models with the Farmwave portfolio to adapt for counting and measuring droplets from sprayers.
RESULTS
More than just changing AI algorithms, we almost built these entirely from scratch given the changes in hardware settings as well. We used both hardware and software zoom factors to combine a method which achieves accuracy and clarity in a moving machine. There is a lot of room for improvement here, and we believe we can achieve better results - possibly to include drift.
KEY LEARNINGS
There’s more to this problem then simply counting and measuring. Benchmark measurements need to be logged and there are many to this. Lighting, spray pattern, application shadowing, boom flex, and camera placement are almost all simultaneously key to this. This activity follows suite to harvesters - asking the question of what to do once you have this info. In addition, if you’re operating a 100ft boom, you’re looking at many nozzle performance checks as once and this level of data capture is going to be too much for any one person to try and keep up with. This requires a large compute capability and delivery mechanism until machinery compute power is updated to handle multi-operational state compute cycles with mechanized changes built in to the protocol.
